RECCR Rensselaer Exploratory Center for Cheminformatics Research
Rensselaer Exploratory Center for Cheminformatics Research (RECCR)












News Members Projects Publications Software Data MLI ECCRS
Support vector regression-based Online Learning Equipment (SOLE)

By Zhuo Zhen
about the developer

The SVR-based Online Learning Equipment (SOLE) is a web-based machine learning system. The SOLE system provides three SVR-based machine learning algorithms that can be used in the Quantitative Structure-Activity Relationship (QSAR)/ Quantitative Structure-Property Relationship (QSPR) studies. For the comparison purpose, this machine learning system also includes Partial Least Squares (PLS) -- a standard linear machine learning algorithm. In addition, several feature selection methods are provided for dimension reduction of the input descriptor space. Finally, two cross-validation methods -- Leave-One-Out (LOO) and K-fold -- are provided for model hyperparameter selection.

Highly suggested: please click and read all the and buttons for new users.

TUTORIAL

\/ Take a quick look at what you can get from this webtool \/

SAMPLE SOLE REPORT PAGES

NOTES
  • This webtool has only been tested on Firefox and Chrome. If you have tested SOLE on anyother browser, please let us know your testing result. (See FEEDBACK on how to give us feedbacks)
  • Please read the tutorial and the "magic buttons" carefully before using. In addition, choose proper parameters and algorithms according to your job.
  • Fuzzy SVR algorithm takes quite a long time, so please be patient.
FEEDBACK
  • We appriciate your feedback for the beta version of this webtool!
  • You may use the feedback button on the bottom right
  • You may email Zhuo Zhen
COMING SOON
  • SAVE MODEL function for future blind test
  • Statistical Diagonostics of the models
Thank you for using SOLE beta version! Please contribute to this work by letting us know your thoughts!

Training and External Test Files

  • Upload Files
    Training File
    Test File
    Triangular Expression
  • Using Sample Files
    Agonist Activity at Human FFA1 Receptor Dataset
    More information please reference here

Feature Selection Method

  • Recursive Feature Elimination
    • Estimator Ordinary Least Squares(OLS)
      Support Vector Machine(SVM)
      Number of Recurtions
      Percentage Removal(/iter)
  • L1-based Feature Selection
    • Parameter C
  • Correlation with Response
    • Remove if less than correlation with Responce
  • Mutual-correlation
    • Mutual-Correlation Threshold
  • Without Feature Selection

Cross-Validation Method

Leave-One-Out (LOO) K-Fold
K

Number of Iterations

Set Random Seed? Yes No

Machine Learning Algorithms

PLS
NO. LV Use Default #LV set
Self Define #LV set

Standard SVR
Kernel Linear
RBF
C Use Default C set
Self Define C set

epsilon Use Default epsilon set
Self Define epsilon set

sigma Use Default sigma set
Self Define sigma set

Fuzzy SVR
Kernel Linear
RBF
C Use Default C set
Self Define C set

epsilon Use Default epsilon set
Self Define epsilon set

sigma Use Default sigma set
Self Define sigma set

LS-SVR
Kernel Linear
RBF
gamma Use Default gamma set
Self Define gamma set

sigma Use Default sigma set
Self Define sigma set



Rensselaer Polytechnic Institute RECCR Home Page || Member Area || Wiki

Copyright ©2009 Rensselaer Polytechnic Institute
All rights reserved worldwide.